Are the Micro-foundations of Money Demand Important?

EC607 is rapidly coming to a close. I’ve finished the RBC model and now I am on to discussing nominal rigidities and New Keynesian Economics. This transition is always somewhat awkward because I have to say something about the demand for money.

Prior to the crisis, money demand had nearly disappeared from mainstream macroeconomics. This might seem strange since so much of macroeconomics involves money, but it’s true. In Advanced Macroeconomics, when money is introduced, Romer simply adds real money balances to the utility function and then moves on to tackle the more important problems of macroeconomics (price rigidities in this case). That’s basically it for money demand: one paragraph and an ad hoc addition to the utility function which is basically never mentioned again.

I think the reason for the marginalization of money demand is two-fold. First, getting money into a neoclassical economic model is really tough. Fiat money (money that isn’t backed by anything with actual value) is simply not valued by market participants in a Walrasian setting. The Walrasian value of something that is intrinsically worthless … is zero. The fictitious Walrasian auctioneer is simply too nimble, too efficient to permit an equilibrium with valued fiat money. To get money in to these models (with any micro-foundations at all) requires that we create some kind of a “gap” in markets to create some room for an unbacked currency.

There are models that do the trick. Often researchers working on money demand use frameworks that are decendents of the Kiyotaki-Wright (1993) matching model. This model imagines that all transactions take place through random matching between individual traders. Because the probability of a “double coincidence of wants” in the Kiyotaki-Wright model is low, (it’s unlikely that you will bump into someone who wants what you have and has what you want) a fiat currency can circulate in equilibrium. More modern versions use an extension suggested by Lagos-Wright (2005) in which traders interact in two sub-periods. During the “day”, there is a centralized market where traders use state-contingent contracts. During the “night” they match in an anonymous trade stage.

These are elegant models and they do capture elements of the motives behind holding money. However, they are not used often by most macroeconomists, who often regard these models as being simply too abstract to be useful. Their abstract nature also makes empirical analysis of these models extremely difficult. (Incidentally, if you are a graduate student looking for a research topic, I would encourage you to look outside of this area. It’s a very difficult area and it doesn’t sell very well on the academic job market. Search and matching in general is very hot right now but “money-search” is not.)

The second reason why money demand has been largely relegated to the sidelines is that there are moderately persuasive arguments that we don’t actually need to understand it to study the macroeconomy – even to study monetary economics itself. The argument goes something like this: The Federal Reserve conducts monetary policy in terms of a nominal interest rate target. Once it decides on the setting for the funds rate it adjusts the money supply to enforce its target. The New Keynesian model is an excellent example of this approach. The simplest NK model has an equation governing the demand for goods and services, an equation governing inflation, and an equation describing the Fed’s operating rule. No mention is made of money supply or demand and many (most?) macroeconomists are perfectly happy with this state of affairs. The possibility that we could avoid the issue of money demand is very attractive – particularly given the difficulties of successfully modeling money demand.

Money demand may be making a comeback though. During the crisis, a lot of concern centered on malfunctioning markets for money-substitutes. Recent work by Arvind Krishnamurthy and Annette Vissing-Jorgensen, Stefan Nagel, and Adi Sunderam emphasize the liquidity aspects of many assets that are not traditionally considered “money.” Treasury bills, Commercial Paper, and highly rated securitized assets all have important liquidity components to their market values. In addition, many people think that the demand for liquid, low risk securities encouraged the creation of more and more securitized subprime loans. Not having a suitable model for money (or money substitutes) seems like a particular shortcoming given recent history.

In 1978, there was an amazing conference at the Federal Reserve Bank of Minneapolis devoted explicitly to the study of micro-foundations of money demand. The papers at the conference were later collected in Models of Monetary Economies.[1] In it, there is an interesting discussion by James Tobin who writes in part 

Why does fiat money … have value? What determines its value? This conference [is] based on two premises. One is that the two questions have not been satisfactorily and rigorously answered. The other is that the answer to the second question […] can be achieved if and only if [we have] a precise answer to the first question […]. I am dubious of both premises.

In hindsight, I think it’s clear that Tobin’s suggestion that we have a satisfactory and rigorous understanding of why people hold money – was at best not entirely correct. His second statement – that we might not need to rigorously understand why people hold money – might be right though my faith in his argument has definitely been shaken by recent events.

[1] This volume is available on line here. The 1978 conference lineup was amazingly good and the manuscript includes among other things, Lucas’ “Pure Currency” model, Townsend’s “Turnpike” model, and an excellent paper by Neil Wallace on money demand in the overlapping generations model. While it doesn’t have any of the modern matching models, it is still an impressive and insightful volume and should be required reading for anyone interested in the pure theory of money.


19 thoughts on “Are the Micro-foundations of Money Demand Important?

  1. Pingback: Are the Micro-foundations of Money Demand Important? – My press. | Edgar's Creative

  2. Pingback: Are the Micro-foundations of Money Demand Important? – My press. | Edgar's Real Edge

  3. This is precisely the right focus! Money demand is the most important macro variable because it places a constraint on both monetary and fiscal policy, the evidence of such constraint being inflation, asset booms and busts and prolonged stagnation at zero lower bound.

    I’m interested in your feedback to this endogenous money ISLM model (link below), which reveals the enormous impact money demand exerts on the business cycle. It suggests that monetary policy should shift focus away from interest rate targeting. Instead, central banks should try to match the monetary base to the demand for asset money. Also in regards to the zero lower bound, it calls for equipping central banks with new tools in the form of consumption and investment credits that can lift time preferences into positive territory thus reducing the demand for asset money. The model also offers a pretty straight forward explanation of the pro-cyclical nature of the gold standard and the Gibson Paradox.

  4. I think we should focus on (near-) money as credit, not money as most liquid asset. I guess we can blame Keynes for the entire liquidity preferences nonsense.

  5. “However, they are not used often by most macroeconomists, who often regard these models as being simply too abstract to be useful.”

    These models are no more abstract than anything else we use. This is just a matter of what you are accustomed to looking at. See the neoclassical growth model for the first time and it might be a puzzler. Your argument seems to be that most people don’t care about this stuff, so we should all feel free to ignore it. Well, as the financial crisis of course made painfully obvious, most of the profession was ignoring monetary and financial phenomena at their peril. Some of us have spent our careers thinking about the details of monetary exchange, financial contracts , and banking, and we think that stuff is useful. It’s your loss if you don’t pay attention.

    • I definitely *like* these models (my personal favorite money demand model is the turnpike model) and I’m definitely not saying that it is necessarily safe to ignore them. It does seem to me that many people (many researchers who work on monetary policy in particular) do ignore these issues and they have a moderately plausible argument for why it might be safe to do so.

      The models also seem (to me) much more abstract than most DSGE models. I don’t know how I would begin to connect, say, the Kiyotaki-Wright triangle model with the data. Same for the circle model and the turnpike model … We could test them in a lab setting but that isn’t really what I have in mind. The day/night model might be somewhat less abstract but even then I’m not sure how I would match it up. I could choose the ratio of day/night time to correspond to the fraction of goods purchased with cash in the economy I suppose. I would guess that this would start to look like a Lucas-Stokey cash/credit model. In your opinion what would you say is the best example of a “useful” endogenous money demand model, meaning that it might be able to be compared with data? Or alternatively, is there an insight from any of the endogenous money demand models that you would think is particularly relevant to the conduct of monetary policy or do you think that the models are still in the development stage?

      • I certainly agree that the basic Kiyotaki-Wright model (indivisible assets and goods), and the Turnpike model are not in a form that we can readily take to central bankers to show them what to do. For models that I think are closer to that, see my AER paper (2012), and my recent working paper on QE. These aren’t at the quantitative stage yet, with aggregate shocks, but that’s not far away.

  6. Thanks for bringing up a fascinating debate. I’ve never understood (and I’ve tried) why many intelligent people say this with a straight face:

    “… there are moderately persuasive arguments that we don’t actually need to understand it to study the macroeconomy – even to study monetary economics itself.”

    Like Groundhog Day, we seemed doomed to repeat economic history, over and over. Fifteen years ago, Neil Wallace wrote this dictum:

    “Money should not be a primitive in monetary theory—in the same way that firm should not be a primitive in industrial organization theory or bond a primitive in finance theory.”

    It’s still a fascinating read (showing that the profession is still struggling with the same questions). It’s available online at the Minneapolis Fed (A Dictum of Monetary Policy, Minneapolis Fed Quarterly Review, Winter 1998).

    • Neil Wallace is one of the primary architects of the classic endogenous theory of money demand (he worked a lot with the OG money demand model in the past and later switched to the Kiyotaki-Wright matching framework later on). Neil’s views on this topic have always been that we need to understand money demand at a deep level if we want to have a real understanding of monetary policy issues. There are still researchers who feel this way but there are many who argue that it is not necessary.

      If the economists who work on the pure theory of money want convince the bulk of the profession that it is important to understand the microfoundations of money demand, they need to be explicit about what this understanding provides. What are the main insights that one gets from the pure theory of money that are necessary to understanding the macroeconomy?

  7. Pingback: [18-4-14] Secular Stagnation | Caótica Economía

  8. Chris House has done some good work in macro. On monetary economics he has the right attitude, but needs to do some reading to brush up on what has been done. He might start with the Williamson and Wright survey in the latest Handbook of Monetary Economics. I am also writing a new survey with Lagos and Rocheteau for JEL and it will be ready very soon.

    In the sort run let me say this: Importantly, he asks “What are the main insights that one gets from the pure theory of money that are necessary to understanding the macroeconomy?” To give some examples, I submit the following Top 10 List. As on David Letterman, or in any finite DP problem, you might want to start at the end, with 10, and work backwards.

    1. To see an application taking search-based theory to the data, look at my Marshall Lecture to the Econometric Society, “Sticky Prices: A New Monetarist Approach” (with Head, Liu & Menzio), published in JEEA (2012). The theory explains sticky nominal prices endogenously; it fits the micro data extremely well; and it shows policy predictions from (Keynesian) models with exogenous price stickiness can be completely wrong. This theory requires search, as it uses Burdett-Judd pricing.

    2. To see a search-and-bargaining model where inflation has much bigger welfare effects than reduced-form papers by Lucas, Cooley-Hansen and many others, look again at Lagos-Wright (2005): our numbers are 10 times bigger. The result comes mainly from using bargaining instead of Walrasian pricing, although search is also relevant quantitatively.

    3. To see a model where optimal policy is different once one models microfoundations properly, see Aruoba and Chugh (JET). In brief, reduced-form MIU models like Chari et al. show homothetic utility implies the optimal (Ramsey) policy entails the Friedman rule i=0, even when other taxes are distortionary; Aruoba-Chugh show that homothetic preferences over goods do not lead to money entering homothetically in the indirect utility (value) function, and this means the optimal policy actually involves i>0 for reasonable parameters. Chari et al. missed this one.

    4. Aruoba and Schorfeide nest New Monetarist and New Keynesian wedges in a fully specified and estimated macro model. Despite reasonably large sticky-price distortions, their estimation results show the best policy is either thr Friedman rule i=0, or very much closer to that than to the prescription that follows from New Keynesian theory. Even with parameter uncertainty, accross several different specifications, the optimum is not what New Keynesian models recommend (0 inflation).

    5. To see a model where once liquidity is modeled endogenously the equity-premium puzzle and risk-free-rate puzzle are not so hard to understand theoretically or empirically, see Lagos (JME). Telyukova and Wright (RES) and Telyukova (RES) do something similar for the credit-card-debt puzzle. There are many other examples addressing apparent puzzles in the macro finance literature.

    6. To see where DIamond and Benabou went wrong by predicting that some inflation can be good because it reduces monopoly power in an economy with sticky prices when money is only a unit of account, look at Craig and Rocheteau (JEEA). They show that once the role of money is made explicit, the inflation distortion dwarfs the DIamond-Benabou effect for any reasonable parameters, and the optimum is actually deflation.

    7. To see how the Phillips curve slopes upward, not downward, in the last 50 years of US data except at very high frequencies, or in a few subperiods like the 60s, see Berentsen et al. (AER). They show how one can account for about half of the US unemployment experience over the period, including stagflation, simply by inflation or nominal interest rate changes in a New Monetarist model incorporating a Mortensen-Pissarides labor market. They argue that search and bargaining are critical to understanding these issues.

    8. To see how New Monetarist models can generate housing bubbles, asset-price cycles and sunspots, and other interesting dynamics in models with reasonable parameter values by introducing liquidity explicitly, see Rocehteau and Wright (JME) or Gu et al. (JPE). This is not the case in models that ignore liquidity considerations.

    9. For a calibrated model with money, capital, bonds, equity, and housing assets that is quantitatively consistent with much of US data at least at medium to longer run frequencies, see “Pledgeability and Liquidity: A New Monetarist Model of Financial and Macro- economic Activity” by me and Venky, in the last NBER Macro Annual. The model can also help us understand financial innovation qualitatively and quantitatively.

    10. See Gu et al. (in progress) “Money and Credit Redux” for illustrations of how changes in credit conditions can be completely neutral once the exchange process and the role of money are made explicit. This paper implies the following note of caution: Any suggestion that credit markets are at the root of any economic problem must rely on redistributive effects. Similar analyses of open market operations and less conventional monetary policy are contained in several recent papers by Williamson including his recent AER paper. This work is very useful for understanding what policy can and cannot do.

    What do you want?

    • “What do you want?” How about empirical evidence?

      The only clear test that I’ve seen is Gorodnichenko and Weber (2013) which tries to determine whether sticky prices are “costly” as measured by financial market returns. In Head et al. mentioned above, endogenous price stickiness arises because firms that don’t change their prices are indifferent between keeping their prices and changing it, i.e. price stickiness is not costly to the firm. But the evidence in GW suggests that price stickiness is in fact quite costly, even showing up in stock return volatility of individual companies after monetary policy news. I’d be delighted to hear your views on this approach.

  9. One version of the last recourse of scoundrel is this: First ask for a reason to take something seriously; then receive 10 solid reasons; then try to pick holes in one aspect concerning one of the 10 reasons while ignoring everything else. Well played.

    If we want to engage in that, however, consider: There is no presumption that the findings by G and W imply causation. A heterogeneous-firm version of Head et al. could attribute these observations to the idea that low quality (productivity) firms endogenously find themselves in the region of strategy space where it is a best response to change prices less frequently. Stock prices are hence correlated with stickiness. Not that I know this would work perfectly, but it may be worth pursuing. In any case, Chris’ reply to my top ten list seems hardly a cause to ignore or dismiss the entire New Monetarist paradigm. And, in any case, do we really think the G and W findings constitute compelling evidence that Keynes and his followers are right? Give me a break.

    “How about empirical evidence?” If one looks at the fit of the Head et al. model to the Klenow-Krystov data, I believe that any reasonable scientist with an open mind to the data should say “This is an interesting new theory consistent with some evidence many others have said was salient. We should pursue the approach and see how it does on other dimensions, and we should flesh out it logical implications in many applications and extensions, both theoretically and empirically.” I would not want to see a novel approach to the issues dismissed because of one study — even if it were “the only clear [conceivable?] test.” This would be like dismissing the Burdett-Mortensen model of the labor market by noting that the simplest baseline version predicts a wage density inconsistent with a stylized fact. Years of work on the Burdett-Mortensen model shows how to accommodate that stylized facts and much much more. This is some of the best work in economics, as it combines solid theory with excellent data and sound econometrics.

    Final thought: A claim to be interested in data and/or policy does not require, nor does it give one license to purse, weak theory — not even for the short run.

    Rock on.

    • Let me try to briefly respond to some of the comments above.

      I’ll start by summarizing the main ideas in the original post:
      • The study of money demand – the “pure theory of money” – has been somewhat marginalized by many mainstream macroeconomists.
      • I suspect that the reasons for this are (1) modelling the pure theory of money is really hard and (2) there is an argument that one doesn’t need to have a foundational understanding of money demand to study macro.
      • The study of money demand might be returning to prominence because of the financial crisis. Money substitutes played a role in many markets during the crisis. Moreover, many of these money substitutes were created by the private sector.

      Let me expand on this a bit

      First, I’m not saying that I necessarily endorse the idea that we don’t need a rigorous understanding of money demand to study macro. I’m just not completely convinced either way.

      Second, a lot of these micro-foundations money models are really difficult to work with. Miguel Molico wrote a paper which tries to get some traction on the endogenous distribution of money holdings in a Kiyotaki-Wright model and, if I remember correctly, it wasn’t what I would call an easy task. My friend Dave Mills did work on trilateral matching in a KW search model and I think that was really difficult too. Of course there are tricks and modelling devices that we can use to simplify matters but there are a lot of micro-money models that just seem daunting.

      Lastly, the new expansion of interest in the study of money substitutes is not just empirical. There is also a substantial set of theoretical papers and many are written by people who were working on search models before the crisis. For instance, Veronica Guerrieri, Rob Shimer and Randy have done work on adverse selection and search which was extended by Guerrieri and Shimer to study endogenous liquidity. There is also a cool paper by Braz Camargo and Ben Lester who use a search framework to analyze how a market that is “frozen” (due to adverse selection in their case) gradually resolves over time and how government programs influence this market. It’s not surprising that these researchers are the first ones to enter this area – they already have the human capital needed to make contributions.

      OK, now some specific replies to earlier comments …

      @Steve Williamson:
      I haven’t read your QJE paper yet. I’ll take a look and let you know if I have comments. I should say that I’m perfectly comfortable with theories that are still “in progress” and currently in a stage that is too early to take to the data. I’m more curious about the implications of the theory which hinge on building a theory of money demand from first principles.

      Your top 10 list is fairly technical. You don’t need to convince me that these are good papers and that their authors are skilled, insightful researchers – I already knew that. I also *like* all of this work. But how do you respond when someone says “I think we will do just fine by plugging in a money-in-the-utility specification.” I’m sure that things will be “different” in another model but are they different in a way that is important? The main reason I personally like the micro-foundations of money demand is that I think the models are really interesting but this doesn’t carry much weight with someone who isn’t as enamored with the models.

      Your point # 5 is really crucial – the equity premium puzzle is certainly due in part to a liquidity premium on safe government bonds. That said, can’t I get even that from a MIU specification?

      • All the papers in my Top 10 list can be used to make the point that microfoundations matter, and that is why I chose them. to repeat one at random, consider number 1:

        1. To see an application taking search-based theory to the data, look at my Marshall Lecture to the Econometric Society, “Sticky Prices: A New Monetarist Approach” (with Head, Liu & Menzio), published in JEEA (2012). The theory explains sticky nominal prices endogenously; it fits the micro data extremely well; and it shows policy predictions from (Keynesian) models with exogenous price stickiness can be completely wrong. This theory requires search, as it uses Burdett-Judd pricing.

        The key idea in that paper is that search-based models with price posting along the lines of Burdett-Judd and Burdett-Mortensen give rise to an endogenous distribution of prices, with profits equal for any price in the support. Combined with money being essential for trade (due to commitment and information frictions) we get sticky nominal prices.

        Where do you see anything like that in a CIA or MIU model? Microfoundations matter.

        I volunteer myself to come to Michigan and give a mini-course on these issues ASAP. For my usual fee.

  10. Pingback: The Neo-Fisherite Rebellion |

  11. Pingback: (18/4/14) | The Econ-Blogosphere

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s